Comparative analysis of different algorithms for VAS station land cover classification with limited training points

The Valencian Anchor Station (VAS) (Spain) is an outstanding site operating as a central location for calibrating and validating numerous remote sensing instruments and products. Hence, an accurate characterization of its land cover is required. This research conducts a land cover classification wit...

Full description

Saved in:
Bibliographic Details
Main Authors: D. García-Rodríguez, A. Pérez-Hoyos, B. Martínez, Pablo Catret Ruber, J. Javier Samper-Zapater, E. López-Baeza, J.J. Martínez Durá
Format: Article
Language:English
Published: Elsevier 2025-05-01
Series:International Journal of Applied Earth Observations and Geoinformation
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1569843225001840
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850135597978484736
author D. García-Rodríguez
A. Pérez-Hoyos
B. Martínez
Pablo Catret Ruber
J. Javier Samper-Zapater
E. López-Baeza
J.J. Martínez Durá
author_facet D. García-Rodríguez
A. Pérez-Hoyos
B. Martínez
Pablo Catret Ruber
J. Javier Samper-Zapater
E. López-Baeza
J.J. Martínez Durá
author_sort D. García-Rodríguez
collection DOAJ
description The Valencian Anchor Station (VAS) (Spain) is an outstanding site operating as a central location for calibrating and validating numerous remote sensing instruments and products. Hence, an accurate characterization of its land cover is required. This research conducts a land cover classification within the VAS station (10 × 10 km2) and its surrounding area (30 × 30 km2) for 2021 using multi-temporal imagery from Sentinel-2 Multispectral Instrument (MSI). Several aspects of land cover classification have been evaluated, including i) the feature selection, ii) the temporal resolution of time series (i.e., monthly, seasonal), iii) the performance of six Machine Learning algorithms (i.e., CART, GTB, k-NN, NB, RF, and SVM, alongside three deep learning models (FC-NN, MLP-ED, and ResCNN) and iv) the optimization of classifier tuning parameters. Furthermore, the study assesses the impact of reducing sample sizes on classifying similar areas, extending the classification to three buffers (1 km, 5 km and 10 km) without increasing reference data. ResCNN emerged as the best-performing model, yielding superior classification metrics (96 % overall accuracy and 95 % kappa score) in July, coinciding with the peak vineyard phenology. Producer’s and user’s accuracy values generally exceeded 90 % for most land cover classes, with some exceptions in more challenging categories such as artificial surfaces and non-irrigated arable land, which showed lower accuracies due to inter-class similarity. Overall, the findings underscore the robust performance of all models in land cover classification, demonstrating the feasibility of achieving high-quality classification with a robust methodology and limited training data.
format Article
id doaj-art-e474880a04a0481082afd7a66d55c699
institution OA Journals
issn 1569-8432
language English
publishDate 2025-05-01
publisher Elsevier
record_format Article
series International Journal of Applied Earth Observations and Geoinformation
spelling doaj-art-e474880a04a0481082afd7a66d55c6992025-08-20T02:31:22ZengElsevierInternational Journal of Applied Earth Observations and Geoinformation1569-84322025-05-0113910453710.1016/j.jag.2025.104537Comparative analysis of different algorithms for VAS station land cover classification with limited training pointsD. García-Rodríguez0A. Pérez-Hoyos1B. Martínez2Pablo Catret Ruber3J. Javier Samper-Zapater4E. López-Baeza5J.J. Martínez Durá6University Research Institute on Robotics and Information and Communication Technologies (IRTIC), Universitat de València, Paterna 46980, Spain; Corresponding author.Albavalor, Science Park Universitat de València, Paterna 46980, SpainEnvironmental Remote Sensing Group (UV-ERS), Universitat de València, 46100 València, SpainUniversity Research Institute on Robotics and Information and Communication Technologies (IRTIC), Universitat de València, Paterna 46980, SpainUniversity Research Institute on Robotics and Information and Communication Technologies (IRTIC), Universitat de València, Paterna 46980, SpainEnvironmental Remote Sensing Group (UV-ERS), Universitat de València, 46100 València, SpainUniversity Research Institute on Robotics and Information and Communication Technologies (IRTIC), Universitat de València, Paterna 46980, SpainThe Valencian Anchor Station (VAS) (Spain) is an outstanding site operating as a central location for calibrating and validating numerous remote sensing instruments and products. Hence, an accurate characterization of its land cover is required. This research conducts a land cover classification within the VAS station (10 × 10 km2) and its surrounding area (30 × 30 km2) for 2021 using multi-temporal imagery from Sentinel-2 Multispectral Instrument (MSI). Several aspects of land cover classification have been evaluated, including i) the feature selection, ii) the temporal resolution of time series (i.e., monthly, seasonal), iii) the performance of six Machine Learning algorithms (i.e., CART, GTB, k-NN, NB, RF, and SVM, alongside three deep learning models (FC-NN, MLP-ED, and ResCNN) and iv) the optimization of classifier tuning parameters. Furthermore, the study assesses the impact of reducing sample sizes on classifying similar areas, extending the classification to three buffers (1 km, 5 km and 10 km) without increasing reference data. ResCNN emerged as the best-performing model, yielding superior classification metrics (96 % overall accuracy and 95 % kappa score) in July, coinciding with the peak vineyard phenology. Producer’s and user’s accuracy values generally exceeded 90 % for most land cover classes, with some exceptions in more challenging categories such as artificial surfaces and non-irrigated arable land, which showed lower accuracies due to inter-class similarity. Overall, the findings underscore the robust performance of all models in land cover classification, demonstrating the feasibility of achieving high-quality classification with a robust methodology and limited training data.http://www.sciencedirect.com/science/article/pii/S1569843225001840Land-cover classificationFeature selectionMachine learningDeep learningMulti-temporal training set size
spellingShingle D. García-Rodríguez
A. Pérez-Hoyos
B. Martínez
Pablo Catret Ruber
J. Javier Samper-Zapater
E. López-Baeza
J.J. Martínez Durá
Comparative analysis of different algorithms for VAS station land cover classification with limited training points
International Journal of Applied Earth Observations and Geoinformation
Land-cover classification
Feature selection
Machine learning
Deep learning
Multi-temporal training set size
title Comparative analysis of different algorithms for VAS station land cover classification with limited training points
title_full Comparative analysis of different algorithms for VAS station land cover classification with limited training points
title_fullStr Comparative analysis of different algorithms for VAS station land cover classification with limited training points
title_full_unstemmed Comparative analysis of different algorithms for VAS station land cover classification with limited training points
title_short Comparative analysis of different algorithms for VAS station land cover classification with limited training points
title_sort comparative analysis of different algorithms for vas station land cover classification with limited training points
topic Land-cover classification
Feature selection
Machine learning
Deep learning
Multi-temporal training set size
url http://www.sciencedirect.com/science/article/pii/S1569843225001840
work_keys_str_mv AT dgarciarodriguez comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT aperezhoyos comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT bmartinez comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT pablocatretruber comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT jjaviersamperzapater comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT elopezbaeza comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints
AT jjmartinezdura comparativeanalysisofdifferentalgorithmsforvasstationlandcoverclassificationwithlimitedtrainingpoints